脸部网格模型中非特征点的自动调整

Adapting non-feature points of facial wire-frame model

模型调整是虚拟人脸合成、基于模型的人脸图象压缩编码的重要步骤。本文提出了一种新的基于径向基函数 RBF(Radial Basis Functions)内插的模型非特征点调整方法 ,其中径向基函数采用逆多二次式。由于 RBF内插具有良好的全局连续性 ,而调整逆多二次式的衰减参数又可保证内插局部性 。

In three dimensional face reconstruction procedures with the model-based method, one of important step is the model adaptation. In this paper, we adapt the non-feature points in the model with radial basis functions (RBF) interpolation according to the known facial feature points, and the inverse multi-quadrics is chosen as the interpolation base functions. We can get a good adaptation result because of the smoothness of RBF interpolation and the reasonable local quality by decreasing the influence of remote feature points to the adapted model point through manipulating the internal factor of inverse multi-quadrics.